Forest Ecology and Management 377 (2016) 83–92

Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage: www.elsevier .com/locate / foreco

Assessing spatial distribution, stand impacts and rate of Ceratocystis fimbriata induced ‘ohi‘a (Metrosideros polymorpha) mortality in a tropical wet forest, Hawai‘i Island, USA

⇑ Corresponding author. E-mail address: leifmortenson@fs.fed.us (L.A. Mortenson).

http://dx.doi.org/10.1016/j.foreco.2016.06.026 0378-1127/Published by Elsevier B.V.

Leif A. Mortenson a,⇑, R. Flint Hughes b, James B. Friday c, Lisa M. Keith d, Jomar M. Barbosa e, Nathanael J. Friday f, Zhanfeng Liu g, Travis G. Sowards b

a USDA Forest Service, Pacific Southwest Research Station, Placerville, CA, USA bUSDA Forest Service, Institute of Pacific Islands Forestry, Hilo, HI, USA c University of Hawai‘i at Ma noa Cooperative Extension Service, Hilo, HI, USA dUSDA Agricultural Research Service, Hilo, HI, USA e Department of Global Ecology, Carnegie Institution for Science, Stanford, CA, USA fDartmouth College, Hanover, NH, USA g Tetra Tech Inc., Contractor for USDA Forest Service, State and Private Forestry, Vallejo, CA, USA

a r t i c l e i n f o a b s t r a c t

Article history: Received 28 March 2016 Received in revised form 11 June 2016 Accepted 15 June 2016 Available online 5 July 2016

Keywords: ‘O hi‘a Metrosideros polymorpha Ceratocystis Hawai‘i Forest pathology Invasion

Pests or pathogens that affect trees have the potential to fundamentally alter forest composition, struc-ture and function. Throughout the last six years, large areas of otherwise healthy ‘ohi‘a (Metrosideros poly-morpha) trees have been dying rapidly (typically within weeks) in lowland tropical wet forest on Hawai‘i Island, USA. This mortality is quite distinct from previous well-documented ‘ohi‘a dieback episodes driven by cohort senescence. Ceratocystis fimbiata was identifed and routinely found associated with rapidly dying individuals of ‘ohi‘a, Hawai‘i’s most widespread native tree. Pathogenicity of this fungus was proven and M. polymorpha was recorded as a new host for C. fimbiata. Mortality of ‘ohi‘a at this scale is of great concern as the understory in these forests is often occupied by invasive non-native plants capable of severely limiting ‘ohi‘a regeneration. Imagery of ‘ohi‘a mortality obtained in 2012 revealed large expanses of greater than expected mortality (i.e., �10%) across 1600 ha. By 2014 ‘ohi‘a mortality levels �10% had spread to 6403 ha, or 30% of total area classifed as ‘ohi‘a in our study area. Further, levels of ‘ohi‘a mortality in feld plots established within the study region averaged 39%, and mortality levels were comparable across size classes and forest compositions. Results from a subset of feld plots re-inventoried one year after plot establishment revealed average annual ‘ohi‘a mortality rates of 24% and 28% based on basal area and stem density measures, respectively; mortality rates were as high as 47% in some feld plots. The dearth of ‘ohi‘a seedling recruitment and characteristic understory dominance of non-native species documented within our research plots, coupled with the lethality of C. fimbriata to ‘ohi‘a, suggest that these forests likely will be dominated by non-native species in the future.

Published by Elsevier B.V.

1. Introduction

Pests or pathogens that affect trees have the potential to funda-mentally alter forest composition, structure and function as readily as fre, deforestation, development, and invasive plant species (Hicke et al., 2012). Globally, disease damage is likely exacerbated by ongoing climate change (Allen et al., 2010). Where diseases affect natural forests, they can have a profound impact on habitat

quality as well as local, regional, and global biodiversity (Hicke et al., 2012; Shearer et al., 2009; Chakraborty et al., 2008). Notable examples of signifcant pathogen-induced forest mortality include Sudden Oak Death caused by Phytophthora ramorum (Rizzo and Garbelotto, 2003), chestnut blight caused by Cryphonectria parasit-ica (Orwig, 2002) and Dutch elm disease caused by Ophiostoma spp. (Walters, 1992).

‘O hi‘a lehua (Metrosideros polymorpha) is arguably Hawai‘i’s most widespread and culturally signifcant native tree (Friday and Herbert, 2006; Mueller-Dombois and Fosberg, 1998). Approx-imately 50 species of Metrosideros occur worldwide, primarily in

84 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

Fig. 1. A typical ‘ohi‘a experiencing symptoms of Ceratocystis fimbriata infection in the Puna and South Hilo Districts of Hawai‘i Island. Tree exhibits one healthy, living branch and one dead branch, a common characteristic of infected trees.

South Pacifc island countries, as well as New Zealand and the Philippines. In Hawai‘i, ‘ohi‘a habitat extends from sea level to tree line at 2800 m, across surface substrates of all ages, and moisture gradients ranging from dry forests to wet forests, where it is dom-inant (Mueller-Dombois and Fosberg, 1998). Collectively, ‘ohi‘a dominated forests occupy ca. 350,000 ha across the Hawaiian archipelago, with 250,000 ha of that total found on Hawaii Island, and the remaining 100,000 found on the other main Hawaiian Islands combined (Gon et al., 2006). Since arriving in Hawai‘i four million years ago (Percy et al., 2008), it has succeeded as a gener-alist, colonizing habitats in almost every environment across the archipelago. Considering all the forested areas including Hawai‘i Island, ‘ohi‘a trees account for approximately 50% of the basal area and 50% of all the stems found within those forests (http://apps.fs. fed.us/fadb-downloads). As a result, ‘ohi‘a forests provide habitat for most of the 35 native forest bird species in Hawaii, 21 of which are threatened or endangered (Mitchell et al., 2005), most of the over 900 species of endemic vascular plants (Wagner et al., 1999), and many of the 5000 endemic species of insects (Miller and Eldredge, 1996). In addition, ‘ohi‘a-dominated forests protect most of the critical watersheds for Hawai‘i’s citizens across the entire state (Hosmer, 1911).

Prior tohumancontact, the rangeof ‘ohi‘a’ forests across themain Hawaiian Islands was much broader than it is today; land conver-sion (forest to agriculture and/or grazing lands), fre, develop-ment, and non-native feral ungulates and widespread invasive plant species establishment have reduced the distribution of ‘ohi‘a’ forest substantially (Cuddihy and Stone, 1990). As the dominant tree species of the majority of forest communities in Hawai‘i, ‘ohi‘a trees bear substantial Hawaiian cultural signifcance. The tree itself was and is considered the physical embodiment of K�u, one of the four principal Hawaiian deities. ‘O hi‘a wood was incorporated into the two most sacred structures of heiau (temples) of governance in Hawai‘i: the ki‘iakua (god fgures) and the lele (offering platform). ‘O hi‘a was a fundamental material for dwellings, tools, weapons, religious structures, food preparation implements and containers. Foliage and fowers continue to be critical components of many of the most valued lei, and tender parts of fowers, foliage and roots have been used in herbal medicine. Prevalent and easily recognized in ecosystems from sea level to tree-line, numerous ‘olelo no‘eau (wise and poetical Hawaiian sayings) refer to ‘ohi‘a trees as positive symbols of strength, sanctity, and beauty (Gon, 2012).

During the last six years, many landowners in the Puna district of Hawai‘i Island have reported, to the University of Hawai‘i agri-cultural extension services, the rapid death (i.e., days to weeks) of previously healthy ‘ohi‘a trees of all size classes, resulting in localized to extensive stands of predominantly dead ‘ohi‘a trees. A similar phenomenon characterized as landscape-scale ‘ohi‘a dieback was observed and thoroughly investigated in the 1970 s and 1980 s (Boehmer et al., 2013; Hodges et al., 1986; Mueller-Dombois, 1987; Mueller-Dombois et al., 2013). A variety of causes were associated with these previous ‘ohi‘a dieback epi-sodes including pathogens such as Phytophthora cinnamomi and Armillaria mellea, and an insect pest, the ‘ohi‘a borer (Plagithmysus bilineatus) (Hodges et al., 1986). Overall, the chief underlying cause was determined to be cohort senescence, or age-driven mortality of even aged stands of ‘ohi‘a that was generally triggered by any number of stressors (Mueller-Dombois et al., 2013). Cohort senes-cence is not uncommon for ‘ohi‘a stands to experience since they often exhibit widespread and synchronous establishment on recently formed lava substrates; stages of senescence are relatively synchronous as well (Mueller-Dombois et al., 2013). Recently observed mortality of ‘ohi‘a individuals and whole stands was unique because it occurred in areas that had not experienced clas-sic ‘ohi‘a dieback as documented in previous dieback research (Jacobi personal communication) and because affected trees were

dying so rapidly. The manner in which trees died was also unique—typically with a branch, fork or tree completely dying before its neighbor showed signs of stress (Fig. 1). In addition, rapid ‘ohi‘a death occurred in and around areas of direct human infu-ence (e.g., homes, gardens, road construction) as well as areas with little or no direct human infuence (e.g., inaccessible wet tropical forest stands far from roads or homes).

In 2014, during exploratory fungal sampling, Ceratocystis fimbri-ata was isolated from dying ‘ohi‘a in an area of recent widespread ‘ohi‘a mortality. Keith et al. (2015) fulflled Koch’s postulates and showed that a particular strain of C. fimbriata is pathogenic to ‘ohi‘a. Elsewhere, Ceratocystis species cause wilt diseases of crops and plantation tree species also planted widely in Hawai‘i; these include sweet potato, coffee, mango, Eucalyptus, and Acacia (Roux et al., 2000). However, there are only records of C. fimbriata causing disease on sweet potato, syngonium, taro, Tongan yam and African tulip in Hawai‘i (Browne and Matsuura, 1941).

In order to begin to address the ecological ramifcations of this new disease to Hawai‘i’s dominant native tree, we posed the following questions: To date, how much area of ‘ohi‘a-dominated forest has experienced C. fimbriata-induced mortality? What is the rate of spread of this disease in areas that we have detected it in? What are the forest dynamics in areas affected by the dis-ease? Do certain size classes of ‘ohi‘a succumb to the disease more readily than others? At what rate are ‘ohi‘a trees dying and do mor-tality rates differ greatly among different stands? To address these questions, we used remotely-sensed imagery to estimate the dis-tribution of C. fimbriata-induced mortality and the change in that distribution between 2012 and 2014. We also measured forest stand composition, structure, and the degree of tree mortality in a series of forest plots established throughout areas experiencing

85 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

the disease. We repeated these measurements in a subset of our forest plots after one year to determine annual rates of tree mortal-ity, occurrence of ‘ohi‘a seedling recruitment, and any other changes in forest composition or structure over time.

2. Methods

2.1. Study region

Our study region included the Puna District and the southern portion of the South Hilo District of Hawai‘i Island, USA (Figs. 2 and 3). The forests of this region are classifed as lowland wet trop-ical forest (Zimmerman et al., 2008). Remotely-sensed imagery was used to map ‘ohi‘a mortality across all landownership, and for-est monitoring plots were located both on State of Hawai‘i man-aged forest reserves (i.e. Malama K�ı, Nanawale, Keauohana and Waiakea) and in small, privately owned forests. Collectively, the forest reserves constitute some of the least altered lowland wet forest in Hawai‘i (Mueller-Dombois and Fosberg, 1998; Hughes and Denslow, 2005; Hughes et al., 2014) as much of this forest type in the districts have been urbanized or converted to agriculture (Cuddihy and Stone, 1990). Primary succession in this region typi-cally manifests as colonization of young, barren lava substrates by a single cohort stand of ‘ohi‘a trees, which grow as a monotypic forest stand during the frst 100–200 years (Atkinson, 1970; Mueller-Dombois and Fosberg, 1998; Hughes et al., 2014). Species diversity and structural complexity increase as additional native

Fig. 2. Change in spatial distribution of ‘ohi‘a mortality �10% from 2012 to 2014 (usingIsland.

tree species (e.g., Diospyros sandwicensis, Psychotria hawaiiensis, Psydrax odorata, and Pandanus tectorius) establish under mature ‘ohi‘a overstory canopies throughout subsequent centuries of suc-cession (i.e., 200–700 years) (Wagner et al., 1999; Zimmerman et al., 2008). Declines in ‘ohi‘a basal area and stem density often occur as dominance by mid-to-late successional species increases (Atkinson, 1970; Zimmerman et al., 2008).

Native vegetation is still found across much of the study region, although signifcant portions are now occupied or dominated by nonnative tree and shrub species, particularly in the understory (Cavaleri, 2014; Hughes et al., 2014; Schulten et al., 2014). Non-native understory species commonly encountered in the study area include Psidium cattleianum (strawberry guava), Clidemia hirta and Melastoma candidum (Zimmerman et al., 2008). Other non-native species in the region were purposefully introduced during nonnative tree planting efforts that took place across Hawai‘i throughout the 1900 s (Nelson, 1965); these include Falcataria moluccana, Casuarina equisetifolia and Eucalyptus robusta (Little and Skolmen, 1989).

Mean annual temperature (MAT) of the study area was 23.8 �C, and mean annual precipitation (MAP) ranged from 2600 mm/year near the coast to 4700 mm/year at elevations approaching 550 m elevation (Giambelluca et al., 2013). The range of elevations and rainfall amounts are typical of ‘ohi‘a forests (Friday and Herbert, 2006). Rainfall for the period 2011–2014 was slightly less than typical, with precipitation being between 60% and 90% of normal average as recorded at stations in Hilo, Pahoa, Kapoho, Keauohana, and Glenwood, all locations adjacent or inside of the affected areas.

remotely sensed images) in the Puna and southern South Hilo Districts of Hawai‘i

86 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

Fig. 3. Distribution of ‘ohi‘a mortality across the Puna and southern South Hilo Districts of Hawai‘i Island determined from remotely sensed images collected in 2014.

The lower rainfall would not be expected to cause signifcant mortality due to drought stress, though, as ‘ohi‘a forests can occur in areas receiving as little as 400 mm precipitation annually. Extre-mely wet ‘ohi‘a forests receive 6300 mm. Soils across the region formed on lava fows emanating from the lower east rift zone of the active K�ılauea volcano, and substrate ages vary from <1 year to ca. 1500 years before present (Moore and Trusdell, 1991). Soils are typically histosols consisting of soil organic matter overlaying relatively unweathered lava. Andisols formed on volcanic ash sub-strates derived from explosive events also occur in limited quantity across the study region (Moore and Trusdell, 1991). Differences in the composition and structure of forests between young substrates

(i.e., <400 YBP) and older substrates (i.e., >400 YBP), which typically contained larger trees, were readily apparent on the ground and in remotely sensed imagery (Hughes et al., 2014). All the soils in the affected area were very young, having developed over the last few centuries. C. fimbriata has not yet reached areas of the island with older, deeply weathered soils.

2.2. Initial investigation and assessment

Mean annual temperature (MAT), mean annual precipitation (MAP) and soil types were determined for the areas examined. In early 2013 we conducted a preliminary inventory of recent ‘ohi‘a

87 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

mortality along all roads in the lower (elevation) Puna district. This was followed by helicopter fights over the same region in the fall of 2013. Because ‘ohi‘a trees died rapidly in our study, leaves turned brown at the same time creating an obvious indicator of recent mortality. However, as wind events capable of removing leaves were variable, fne branches were used as indicators of recent mortality as they typically remain for at least two years. We generated an initial estimate of the spatial distribution of recent ‘ohi‘a mortality based on these efforts. This distribution layer was combined with layers of spatially explicit MAP, roads and residential development, and geologic features such as lava fow boundaries and fault zones in ArcGIS (ESRI Inc., Redlands, CA, USA).

2.3. Spatial distribution and temporal change of ‘ohi‘a mortality

Using aerial photographs from February 2012 at 15 cm of spa-tial resolution (Pictometry International Inc. Henrietta, New York, USA), we mapped ‘ohi‘a mortality within an area of 33,133 ha. Prior to mapping, an extensive visual training in areal signature of dead and live ‘ohi‘a, based on its color, size and shape, was performed on known ‘ohi‘a/non-‘ohi‘a locations throughout the Puna district. Subsequently, the imagery was partitioned into 100 � 100 m grid cells, and we visually inspected the imagery within each grid cell to assign three classes: ‘‘Non O hi‘a Forest”, ‘‘‘O hi‘a no mortality”, and ‘‘‘O hi‘a mortality �10%”. In 2013 we were not aware of the potential large extent of this novel ‘ohi‘a disease (C. fimbriata), and as such, our feld validation data were restricted to the three above-mentioned classes. ‘‘Non O hi‘a Forest” class refers to all other land cover types, such as urban areas and agricultural lands. The ‘‘‘O hi‘a no mortality” class includes forest stands where ‘ohi‘a dead trees were not found. The third class, ‘‘‘O hi‘a forest mortality �10%”, includes forest stands where ‘ohi‘a dead trees represent �10% of each 100 � 100 m grid cell. We assessed the 2012 map accuracy using 34 feld validation points collected in 2013, when the study commenced (estimates of how forest conditions would have appeared in 2012 were generated), and computing the Cohen’s Kappa coeffcient (Congalton, 1991) and overall disagree-ment (see details in Pontius and Millones, 2011; Pontius and Santacruz, 2014). The overall disagreement is the sum of the quan-tity disagreement (amount of disagreement between reference and classifcation data) and the allocation disagreement (amount of disagreement between reference and classifcation data that is due to the less than optimal match in the spatial allocation of the categories, given the proportions of the categories). We calcu-lated these accuracy metrics using a contingency table. Although the number of feld validation points is restricted, mapping errors may be minimized by the high spatial resolution of the images, our extensive effort in classifying the imagery, and the low diversity of canopy species found in the study area.

We re-evaluated ‘ohi‘a mortality, across the same region in 2014 using World View-2 satellite imagery (8-band multispectral data) at a 50 cm spatial resolution collected in March of that year. In 2014, our feld data indicated that patch sizes and severity of ‘ohi‘a mortality within forest stands had greatly intensifed, and at this time we examined and quantifed gradients of mortality severity. As such, the 2014 imagery was also partitioned into 100 m � 100 m grid cells, and each grid cell was evaluated and assigned an ‘ohi‘a mortality severity rating to provide collective estimates of the distribution and severity of ‘ohi‘a mortality across the study region (Oblinger et al., 2012). ‘O hi‘a mortality in all grid cells was visually classifed; i.e., each grid cell was manually exam-ined to determine whether ‘ohi‘a was present in the cell and, if so, what percentage were dead. Cells were classifed as high severity (�50% of ‘ohi‘a dead), above expected (10–49.9% of ‘ohi‘a dead), noticeable (0.1–9.9% ‘ohi‘a dead) and no ‘ohi‘a mortality. The areal

extent of ‘ohi‘a forest in our study region was obtained from state HIGAP GIS layers (USGS 2001). For 2014 imagery, we carried out ground verifcation in 40 points randomly selected along road-sides and classifed in the feld regarding ‘ohi‘a mortality severity and ‘ohi‘a forest type. These feld assessments were applied to our ‘ohi‘a mortality mapping layers to produce a contingency table that included the classes: high, above-expected, noticeable, ‘ohi‘a forest-no mortality, non ohi‘a forest. To perform a temporal comparison between 2012 and 2014, we also reclassifed the 2014 map into three classes: ‘‘non ohi‘a Forest”, ‘‘‘ohi‘a no mortality”, and ‘‘‘ohi‘a mortality �10%”. We then calculated Cohen’s Kappa coeffcient, quantity disagreement, allocation disagreement, and overall dis-agreement (Pontius and Millones, 2011; Pontius and Santacruz, 2014) to assess map accuracy for all map categories. In both 2012 and 2014 maps, we defned ‘ohi‘a mortality based on dead trees from any potential cause, such as senescence, weather events or anthropogenic infuence. As such, it’s probable that some ‘ohi‘a mortality in our study region was not due to C. fimbriata. Finally, we calculated the total temporal change in area of ‘O hi‘a mortality �10% by overlapping the 2012 and 2014 maps.

2.4. ‘O hi‘a mortality monitoring plots

We established an initial set of monitoring plots (n = 9) in areas of recent ‘ohi‘a mortality or adjacent to such areas in the lower Puna district during January and February of 2014; plots were established to document both recent and future ‘ohi‘a mortality. Plot location were selected to capture variety with regard to abun-dance of non-native species, composition, stem density, proximity to anthropogenic activities, and prior recent ‘ohi‘a mortality levels. These plots were re-inventoried in the winter of 2015 to document ‘ohi‘a mortality rates, and a new set of monitoring plots was estab-lished at this time as well. A new set of plots (n = 9) were estab-lished where C. fimbriata infection of ‘ohi‘a trees was confrmed in order to document future ‘ohi‘a mortality and stand changes, rather than document past ‘ohi‘a mortality. Data from this second set of monitoring plots was not included in analyses of ‘ohi‘a mortality rates. Many of the second set of monitoring plots were established in the South Hilo district, which lies north of the lower Puna district, because C. fimbriata infection had been confrmed in these areas at the time of 2014 plot establishment. All plots were separated by �100 m.

Sampling and measurements were conducted on GPS fxed-radius 18 m, permanently monumented, circular plots, with 6 m radius co-located subplots to measure invasive species presence/ abundance and ‘ohi‘a regeneration (all seedlings �15 cm in height inventoried). A vegetation profle was conducted on a 6 m radius subplot within each plot to provide an ocular estimation of percent cover and mean height by species (see USDA Forest Service, 2013). Each tree �1.5 cm DBH encountered in the plot was tallied, and its species identity, DBH, crown position [i.e. open grown, dominant, co-dominant, suppressed, overtopped (USDA Forest Service, 2013)], and status (live or dead) were recorded. The presence of C. fimbriata in ‘ohi‘a trees undergoing mortality was confrmed by analysis of wood samples of symptomatic trees (e.g., attached intact brown leaves on selected limbs or the entire canopy) located within 200 m of all monitoring plot locations. Pathogen presence was confrmed in each wood sample via analysis at the USDA, Agricultural Research Service, Pacifc Basin Agricultural Research Center (PBARC), Hawai‘i, USA, using techniques described by Keith et al. (2015).

2.5. Statistical analysis of ‘ohi‘a mortality monitoring plots

A generalized linear mixed model analysis was executed in order to calculate the probability of individual ‘ohi‘a trees being

88 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

dead using a cubic polynomial of the natural log of DBH size, after accounting for the variation in mortality among plots (SAS, SAS Institute Inc. Cary, NC, USA). Annual mortality rates were calculated as a simple percentage of ‘ohi‘a stems �1.5 cm DBH live in 2014 that were dead in 2015 (re-inventoried at same time of year). We also calculated 2014–2015 mortality by% basal area lost (dead). Spearman rank order correlation tests (P � 0.05) of P. cattleianum abundance and basal area on arcsine-transformed ‘ohi‘a mortality rate (winter 2014 to winter 2015) were conducted for each plot (SigmaPlot, Systat Software Inc. Chicago, IL, USA).

3. Results

3.1. Spatial distribution of ‘ohi‘a mortality

Analysis of remotely-sensed imagery indicated that areas exhibiting substantial ‘ohi‘a tree mortality (i.e., �10% of ‘ohi‘a trees dead) increased from 1600 ha in 2012 to 6403 ha in 2014. This increase represented 30% of the total area classifed as ‘ohi‘a forest in our study region in winter 2014 (Fig. 2). We highlight that the classifcation of ‘ohi‘a mortality �10% includes a large range of mortality, mainly in the 2014 mapping (see Fig. 3). High severity mortality (�50% of ‘ohi‘a trees dead) was found in 11% of the total

Table 1 ‘O hi‘a mortality severity mapped from 2014 World View II satellite imagery.

Hectares % of ‘ohi‘a forest in study area

High severity mortality (red in Fig. 3) 2419 11.2% �50% ‘ohi‘a dead

Above expected mortality (orange in Fig. 3) 3984 18.5% 10–49.9% ‘ohi‘a dead

Noticeable mortality (Light green in Fig. 3) 3824 17.7% 0.1–9.9% ‘ohi‘a dead

‘O hi‘a forest, no mortality (green in Fig. 3) 11,341 52.6%

Table 2 Stem density (stems ha�1), basal area (BA, m2 ha�1), and percent mortality of native trees snative shrub percent cover in forest plots established in Hawaiian forests undergoing Rap

Elevation Substrate ‘O hi‘a Psychotria Diospyros age hawaiiensis sandwicensis

Plot (m) (YBP) Density BA Density BA Density BA

Established 2014 1 312 400–750 904 39.9 88 0.7 0 0.0 2 160 750–1500 305 13.7 0 0.0 0 0.0 3 291 400–750 560 38.7 49 0.5 0 0.0 4 287 400–750 747 22.6 0 0.0 0 0.0 5 274 400–750 599 34.5 167 1.1 128 2.3 6 285 400–750 462 62.4 855 3.9 0 0.0 7 232 400–750 422 42.6 236 2.8 275 2.3 8 185 400–750 1700 40.6 5109 12.2 0 0.0 9 179 400–750 275 23.9 20 0.1 29 0.1

Established 2015 10 284 400–750 1238 44.9 108 0.6 0 0.0 11 253 750–1500 973 14.3 20 0.0 20 0.2 12 299 750–1500 314 29.7 10 0.2 0 0.0 13 321 750–1500 403 20.8 69 0.3 0 0.0 14 342 750–1500 255 19.1 20 0.6 0 0.0 15 398 750–1500 737 38.4 324 1.0 0 0.0 16 240 200–400 894 14.4 20 0.1 0 0.0 17 242 400–750 88 21.5 334 8.3 432 8.1 18 112 200–400 2004 10.3 0 0.0 0 0.0

Mean 715.5 29.6 412.6 1.8 49.1 0.7

a Nearly 100% of ‘O hi‘a trees were dead at installation; data from plot 9 was not inclu

area classifed as ‘ohi‘a forest in our study region in winter 2014 (Table 1 and Fig. 3). An additional 3824 ha exhibited noticeable (visible mortality <10%) ‘ohi‘a mortality (Table 1 and Fig. 3). Taken together, 47% of area classifed as ‘ohi‘a forest in our study region exhibited visible ‘ohi‘a mortality. The 2014 imagery showed recent ‘ohi‘a mortality occurring in the South Hilo district (Fig. 2), where it had not been present in the 2012 imagery. The 2012 mapping showed an overall disagreement of 12% (quantity disagreement of 10%; allocation disagreement of 2%, and kappa of 0.77). The 2014 mapping showed different accuracies as we performed classifcation using three (Fig. 2) or fve classes (Fig. 3). Using three classes, the 2014 map showed an overall disagreement of 18% (quantity disagreement of 8%; allocation disagreement of 10%, and kappa of 0.76). Using fve classes the 2014 map showed an overall disagreement of 30% (quantity disagreement of 10%; alloca-tion disagreement of 20%, and kappa of 0.61). While analyzing remotely sensed data, we observed a trend that most ‘ohi‘a mortal-ity documented in this study occurred primarily on lava fows aged 400–750 year before present (YBP). Additionally, our preliminary ground inventory revealed mortality of many ‘ohi‘a trees growing on areas of lava fows <400 YBP that were adjacent to �400 YBP fows.

3.2. Stand characteristics

‘O hi‘a was the dominant native woody species in all of the 18 forest plots; on average, it comprised 58% ± 7% of total native stem density (�1.5 cm DBH) (i.e., stems ha�1) and 89% ± 2% of total native tree basal area (�1.5 cm DBH) (i.e., m2 ha�1) (Table 2). In contrast, the two next most common native trees, P. hawaiiensis (kopiko) and D. sandwicensis (lama) collectively accounted for an average of 38% ± 7% of total native stem density and 8% ± 3% of total native basal area (Table 2). P. cattleianum, the most common non-native woody species across our study plots, exhibited values that were on average 35% of the average total native woody basal

pecies and a common non-native tree (Psidium cattleianum), as well as native and non-id ‘O hi‘a Death across the Puna and South Hilo Districts of Hawai‘i Island.

All native Psidium Native Non- ‘O hi‘a mortality trees cattleianum shrub native 2014–2015

cover shrub cover

Density BA Density BA % % (% of (% of stems) BA)

1100 42.7 3979 15.4 0 6 50% 49% 305 13.7 0 0.0 65 20 42% 35% 609 39.2 12,556 41.5 0 0 46% 54% 747 22.6 7692 29.5 0 10 25% 7% 894 37.9 11,671 27.8 0 4 13% 8% 1434 68.7 0 0.0 4 68 23% 9% 933 47.7 7339 18.6 0 0 3% 3% 6808 52.9 5924 6.5 0 89 25% 29% 432 27.0 0 0.0 3 95 a a

1444 47.2 15,650 30.5 3 0 1022 14.7 7692 12.4 83 9 403 31.9 265 0.7 90 9 501 21.9 1149 3.5 95 4 305 22.1 88 0.2 96 2 1385 45.0 3979 10.3 1 19 914 14.5 0 0.0 99 0 865 38.0 7604 8.1 4 3 2004 10.3 3360 1.3 65 5

1228.0 33.2 4941.7 11.4 33.8 19.1

ded in annual mortality rate calculations.

89 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

area, and stem density values that were fve times greater average total native stem densities (Table 2).

All ‘ohi‘a size classes exhibited substantial levels of mortality prior to the measurement of each plot (Fig. 4). On average, 39% ± 7% of ‘ohi‘a stem density, and 39% ± 7% of ‘ohi‘a basal area, had died in plots at the time of initial plot establishment in 2014 and 2015 (Table 3). In contrast, P. hawaiiensis, a common native understory tree, exhibited only 7% mortality at the time of plot installation (data not shown). Less common native species, Mysine lessertiana and D. sandwicensis, did not exhibit higher than expected mortality (i.e., �10% dead), and less than 1% of the non-native P. cattleianum stems inventoried exhibited mortality (data not shown).

3.3. ‘O hi‘a mortality rates from 2014 to 2015

Between 2014 and 2015, annual mortality rates of ‘ohi‘a in re-inventoried monitoring plots averaged 28% ± 6% when expressed as a percentage of total stems, and 24% ± 7% when expressed as a percentage of total basal area (Table 2). Mortality rates varied among individual plots; stem density and basal area

1.0

0.8

0.6

0.4

0.2

0.0

Mor

talit

y

0 20 40D

Fig. 4. A generalized linear mixed model showing the probability of all individual ‘ohivariation in mortality among plots.

Table 3 Stem density, basal are, and percentage mortality of ‘ohi‘a in trees encountered in forest mand South Hilo Districts Hawai‘i Island, USA.

‘O hi‘a density (stems ha�1) Mortality at installment ‘O hi‘a BA

2014 2014 2015 2015 2014 2

Plot # Live Dead Live Dead (% of stems) Live D

Installed 2014 1 157 747 79 825 83% 8.8 32 236 69 138 167 23% 9.6 43 275 285 147 413 51% 18.1 24 118 629 89 658 84% 3.7 15 452 147 393 206 25% 27.0 76 128 334 99 363 72% 18.3 47 383 39 373 49 9% 40.1 28 815 884 609 1090 52% 16.0 29 10 265 10 265 96% 0.05 2

Installed 2015 Mean ± S.E. 10 – – 737 501 40% – –11 – – 776 196 20% – –12 – – 275 39 13% – –13 – – 324 79 20% – –14 – – 138 118 46% – –15 – – 540 196 27% – –16 – – 776 118 13% – –17 – – 79 10 11% – –18 – – 1768 236 12% – –

Mean + S.E. 39 ± 7%

a Plot containing primarily dead ‘ohi‘a at installment; data from plot 9 was not includ

values were as high as 50% and 54%, respectively, and as low as 3% in each case (Table 3). We did not include plot # 9 in this por-tion of our analysis because virtually all ‘ohi‘a trees within the plot had suffered Ceratocystis-induced mortality prior to plot establish-ment (Table 3). The lack of clear C. fimbriata-free areas prohibited the successful use of controls within our study region, although we had an annual (stem and basal area) mortality rate of 3% on plot 7 [our only plot that showed no evidence of decline from 2014 to 2015 (despite testing positive for C. fimbriata,)] (Table 3).

3.4. ‘O hi‘a regeneration, invasive species pressure and vegetative composition

We encountered only 13 ‘ohi‘a seedlings across the 18 monitor-ing plots (1.8 ha), and 11 of those 13 seedlings were located in a single monitoring plot that exhibited high native species diversity and minimal recent ‘ohi‘a mortality (Table 2). Further, we found no correlation between P. cattleianum stem density (rs = 0.285, P = 0.434) or basal area (rs = 0.119, P = 0.742) and ‘ohi‘a mortality rate (high P values in Spearman rank order correlation tests indicate no correlation). Total shrub cover averaged 53% ± 10% for

60 80

BH (cm)

‘a trees encountered in feld monitoring plots being dead, after accounting for the

onitoring plots located within areas experiencing Rapid ‘O hi‘a Death across the Puna

(m2 ha�1) Mortality at installment Mortality

014 2015 2015 2014–2015

ead Live Dead (% of BA) (% of stems) (% of BA)

1.1 4.5 35.4 78% 50% 49% .1 6.2 7.5 30% 42% 35% 0.5 8.2 30.4 53% 46% 54% 8.9 3.5 19.1 83% 25% 7% .5 24.7 9.8 22% 13% 8% 4.2 16.7 45.8 71% 23% 9% .5 39 3.6 6% 3% 3% 4.6 11.2 29.4 61% 25% 29% 3.9 0.05 23.9 99.8% a a

28% ± 6% 24% ± 7% 38.1 6.8 15% – – 12.5 1.8 13% – – 27.7 2.0 7% – – 17.3 3.5 17% – – 7.6 11.5 60% – – 31.4 7.1 18% – – 12.0 2.5 17% – – 17.9 3.6 17% – – 7.3 2.9 29% – –

39 ± 7%

ed in annual mortality rate calculations.

90 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

all plots, and non-native shrub cover averaged 19% ± 7%. Common species included the native fern, Dicranopteris linearis (uluhe) whose cover averaged 33% across all plots, and non-native shrubs, C. hirta and M. candidum, whose cover averaged 7% and 4% across all plots, respectively (data not shown).

4. Discussion

Our results document a pronounced stand- and landscape-scale manifestation of Ceratocystis-induced mortality (commonly referred to as Rapid ‘O hi‘a Death, or ROD in Hawai‘i). This current widespread ‘ohi‘a mortality has occurred in an entirely different region than past ‘ohi‘a dieback events (Mueller-Dombois et al., 2013), and our results indicate that C. fimbriata, a primary patho-gen of ‘ohi‘a, is widespread across current affected areas and present in or adjacent to our plots (Keith et al., 2015). Our results clearly documented Ceratocystis-induced mortality of ‘ohi‘a across all size classes and competitive position (i.e. levels of suppression or dominance). This pattern is quite distinct from the cohort senes-cence/dieback phenomenon described by Mueller-Dombois et al. (2013) where mortality occurred in older trees. Keith et al. (2015) demonstrated C. fimbriata to be consistently lethal to ‘o hi‘a seedlings and saplings growing in ideal conditions of labora-tory growth chambers. Furthermore, P. cattleianum abundance (and basal area) was not correlated with ‘ohi‘a mortality rates, suggesting that competition from invasive, non-native species was not a predisposing factor for C. fimbriata-induced mortality. Further, no other co-occurring species in our plots exhibited elevated levels of mortality or symptoms of C. fimbriata infection.

Our results clearly illustrate the dominance of ‘ohi‘a with regard to composition and structure, of the native wet lowland forests of the Puna District of Hawai‘i Island. That ‘ohi‘a trees accounted for 72% of total native tree density and 91% of total native basal area in forest plots measured here is in close agreement with prior stud-ies conducted in the same region. Across a series of forest types in lowland wet forests of Hawaii, Atkinson (1970) concluded that suc-cession to ‘ohi‘a dominated forest was the most general trend. Hughes et al. (2014) documented dominance of ‘ohi‘a stands using LiDAR data across an inventory of 1300 ha of native forests varying widely with respect to forest age as well as substrate age and type (i.e., pahoehoe or a’a lava). Zimmerman et al. (2008) characterized species composition and forest structure in plots totaling 9 ha of sampling area across forest stands that varied by lava age and type. They noted that although ‘ohi‘a was clearly the most dominant native tree on both young and old lava substrates sampled in their study, very little ‘ohi‘a seedling recruitment was observed, and non-native invasive species – notably P. cattleianum and M. candidum - were dominant understory components beneath ‘ohi‘a overstory canopies, particularly on substrates >300 years in age (Zimmerman et al., 2008). These authors concluded, along with Hughes et al. (2014) and Atkinson (1970) that widespread mortal-ity of ‘ohi‘a trees by whatever means will very likely transform these forests from native dominated to non-native dominated. As such, the current phenomenon of Ceratocystis-induced ‘ohi‘a mor-tality is of utmost concern because it both hastens the loss of the compositional, structural, and functional component of these native forests and paves the way for dominance of invasive, non-native plant species across large expanses of these wet lowland forest ecosystems. The potential loss of ‘ohi‘a is crucial, not just because of its dominance in these lowland forests, but because of its dominance throughout the complete history of these lowland forests, and its dominance across the Hawaiian landscape.

Because we mapped mortality of all dead ‘ohi‘a trees, regardless of cause in our study area, it’s inevitable that some ‘ohi‘a individu-als in our study area have died from anthropogenic or natural

causes other than C. fimbriata. The pathogenic rust fungus Puccinia psidii has also become more widespread in ‘ohi‘a across Hawai‘i (Uchida et al., 2006), yet it does not appear capable of killing mature trees (Silva et al., 2014), although the disease does kills ‘o hi‘a seedlings in some locations within our study region. Our accuracy assessment for 2012 and 2014 imagery mapping indicate substantial agreement with feld data (Landis and Koch, 1977; Congalton, 1991; Pontius and Millones, 2011), suggesting some error may be present in our mapping of ‘ohi‘a mortality. However, the increase in area affected and severity of ‘ohi‘a mortality is undeniable as both plot and remote sensing data indicated similar levels of disease spread.

Due to its extreme ecological and cultural importance as a native Hawaiian species, previous episodes of ‘ohi‘a forest decline on Hawai‘i Island have been studied in depth (Mueller-Dombois et al., 2013). Following nearly 40 years of research, cohort senes-cence was identifed as the chief determinant of landscape level ‘ohi‘a dieback (Mueller-Dombois et al., 2013). Boehmer et al. (2013) recently noted greater ‘ohi‘a regeneration in areas having experienced cohort senescence compared to areas that had not experienced cohort senescence. Such fndings suggest an overall stable long-term ‘ohi‘a population despite the occurrence of past severe dieback episodes (Hodges et al., 1986). However, none of these past dieback episodes has been driven by disease or insect outbreaks (Hodges et al., 1986) and none has occurred in areas of high non-native plant abundance seen in our study region (Zimmerman et al., 2008; Mascaro et al., 2008; Cavaleri, 2014).

Extensive C. fimbriata-induced mortality in natural forest sys-tems is rare. The few documented cases elsewhere in the world include mortality of the riparian species Platanus orientalis in Greece (Ocasio-Morales et al., 2007) and mortality of Prosopis cineraria and Dalbergia sissoo in Oman and Pakistan (Al Adawi et al., 2013; Poussio et al., 2010). Mortality levels as high as 80% along canal banks and 40% along roads were reported in investiga-tions of D. sissoo decline in Pakistan (Bajwa et al., 2003), although other pathogens such as Fusarium solani and P. cinnamomi were found to be present in addition to C. fimbriata s.l. (Poussio et al., 2010). In Greece, C. fimbriata-caused mortality of P. orientalis occurred primarily along streams and rivers and was limited to a 400 km2 area (Ocasio-Morales et al., 2007). C. fimbriata-induced mortality has also been documented in aspen trees (Populus tremu-loides) throughout USA and Canada, but with substantially lower mortality rates than we documented in ‘ohi‘a (Hinds, 1972). In most of the aforementioned cases, spread of C. fimbriata via root grafting was hypothesized (Ocasio-Morales et al., 2007). Root con-tact may offer a mechanism for spread of C. fimbriata in Hawai‘i as well; spread via waterways may not be as important, as our study region lacks perennial streams and rivers. With consistent mechanical injury lacking in the natural forest, wounding via wind or animal could also offer a mechanism of spread. There has been no documentation of a new invasive insects in the area but spread through native insects such as P. bilineatus remains a possibility. C. fimbriata was confrmed in numerous dead and dying ‘ohi‘a across all ages of lava fows in the region by winter 2015 (L. Keith, per-sonal communication), yet, based on coarse fow age maps, the majority of documented recent ‘ohi‘a mortality in this study occurred on lava fows aged 400–750 YBP. Additionally, we consis-tently observed this pattern in the feld, and suggest it may offer clues about potential contributing factors or C. fimbriata vectors, yet we were unable to test it to a level of satisfactory robustness. Further, our results documented ‘ohi‘a mortality as occurring both in areas near human infuence as well as in areas away from direct human contact.

Rates of annual mortality at the stand level and rates of increase in the spread of mortality at the landscape-scale reported here were greater than other instances of C. fimbriata induced mortality

91 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

anywhere in the world (Bajwa et al., 2003; Poussio et al., 2010; Ocasio-Morales et al., 2007). Though a thorough estimate of ‘ohi‘a background mortality is lacking, annual background mortal-ity in tropical wet rainforest globally was considered to be <5% by Lugo and Scatena (1996), considerably lower than the annual rates we report. Furthermore, in 2004 Hughes and Denslow (2005) found 97% of total ‘ohi‘a basal area to be live during their study in healthy ‘ohi‘a forest in the Puna District. We were unable to establish control plots as no suitable locations in our study area were clearly C. fimbriata free. The comprehensive ‘ohi‘a mortality we are currently observing in Hawai‘i appears comparable to the introduced Dutch elm disease and chestnut blight on the US main-land. Over half the elm trees in the northern US have succumbed to Dutch elm disease (Stack et al., 2011), and the chestnut blight killed an estimated 3.5 billion American chestnut trees in the east-ern U.S. between 1904 and 1940 (Raupp et al., 2006). Given results from this study, it is not unreasonable to expect analogous levels of Ceratocystis-induced ‘ohi‘a mortality over time in Hawai‘i.

Although C. fimbriata-induced mortality has not been documented in naturally occurring wet forests of the tropics to the extent documented here, numerous studies document C. fimbriata-induced mortality in plantation settings around the world. Plantations of E. grandis in South America, Africa and Australia have experienced severe Ceratocystis mortality (Roux et al., 2000; Barnes et al., 2003). Ceratocystis species have had devas-tating effects on Acacia mearnsii in South Africa (Roux and Wingfeld, 2009). Recently, Ceratocystis has caused wilt symptoms and rapid mortality in Acacia mangium plantations across much of Indonesia and Vietnam (Tarigan et al., 2011; Thu et al., 2012). In Vietnam, annual mortality rates were 5–7% in plantations (Thu et al., 2012), values lower than annual ‘ohi‘a mortality rates reported here. Ceratocystis fagacearum-induced oak wilt disease is now present across 24 U.S. states (Juzwik et al., 2011). Ceratocystis spp. infection in plantations often enters individual trees through wounds generated by pruning or other mechanical damage. Addi-tionally, spores of many Ceratocystis species survive for extended periods in soil and can be spread via movement of contaminated soil (Marin et al., 2003).

Increases in the extent of ‘ohi‘a mortality (defned as �10% mor-tality of ‘ohi‘a), coupled with a high rate of annual ‘ohi‘a mortality, indicate that C. fimbriata induced mortality occurs quickly and spreads rapidly. Mean annual plot mortality rates from 2014 to 2015 suggest that ‘ohi‘a mortality may be nearly eliminated in as little as four years in particular areas. These mortality rates are substantially higher than the 5.4% annual mortality rate noted by McPherson et al. (2010) for Notholithocarpus densiflorus experienc-ing Sudden Oak Death (SOD) in California, a disease which has killed more than >5 million trees along the coasts of Washington and Oregon over the last 20 years (Frankel and Palmieri, 2014). In addition, the failure of ‘ohi‘a seedling recruitment and character-istic understory dominance of non-native species documented within our research plots, coupled with the lethality of C. fimbriata to ‘ohi‘a (Keith et al., 2015), suggest that these forests will be dom-inated by non-native species in the future. On a more positive note, ‘ohi‘a exhibits a high degree of genetic variability (Stacy et al., 2014) which may give some ‘ohi‘a individuals predisposition to survive C. fimbriata in a manner similar to survivorship exhibited by Quercus agrifolia individuals exposed to Ceratocystis fagacearium infection (Dodd et al., 2005). Such patterns remain to be seen.

Acknowledgements

We sincerely thank P. Cannon and J. Jacobi for their guidance and assistance with imagery acquisition. We are also grateful for feld and laboratory assistance provided by L. Sugiyama, K. Dunn, B. Luiz and B. Boche. We thank P. Foulk for help with data collation

and editing. We thank J. Baldwin for assistance with statistical analysis. We are grateful for constructive manuscript reviews by C. Fettig and anonymous reviewers. We thank Region 5 State and Private Forestry of the USDA Forest Service and the Pacifc Southwest Research Station of USDA Forest Service for substantial funding of this research.

References

Al Adawi, A.O., Barnes, I., Khan, I.A., Al Subhi, A.M., Al Jahwari, A.A., Deadman, M.L., Wingfeld, B.D., Wingfeld, M.J., 2013. Ceratocystis manginecans associated with a serious wilt disease of two native legume trees in Oman and Pakistan. Australas. Plant Pathol. 42 (2), 179–193.

Allen, C.D., Macalady, A.K., Chenchouni, H., Bachelet, D., McDowell, N., Vennetier, M., Kitzberger, T., et al., 2010. A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. For. Ecol. Manage. 259 (4), 660–684.

Atkinson, I.A.E., 1970. Successional trends in coastal and lowland forest of Mauna Loa and Kilauea Volcanoes, Hawaii. Pac. Sci. 24, 387–397.

Bajwa, R., Javid, A., Shah, M.B.M., 2003. Extent of Shisham (Dalbergia sissoo Roxb.) decline in Sialkot, Gujranwala, Lahore and Sargodha Districts. Mycopathology 1, 1–5.

Barnes, I., Roux, J., Wingfeld, B.D., O’Neill, M., Wingfeld, M.J., 2003. Ceratocystis fimbriata infecting Eucalyptus grandis in Uruguay. Australas. Plant Pathol. 32, 361–366.

Boehmer, H.J., Wagner, H.H., Jacobi, J.D., Gerrish, G.C., Mueller-Dombois, D., 2013. Rebuilding after collapse: evidence for long-term cohort dynamics in the native Hawaiian rain forest. J. Veg. Sci. 24 (4), 639–650.

Browne, A.C., Matsuura, M., 1941. Black rot of sweet potato. In: Agricultural Extension Circular 134. University of Hawaii, p. 2.

Cavaleri, M.A., Ostertag, R., Cordell, S., Sack, L., 2014. Native trees show conservative water use relative to invasive trees: results from a removal experiment in a Hawaiian wet forest. Conserv. Physiol. 2 (1), cou016.

Chakraborty, S., Luck, J., Hollaway, G., Freeman, A., Norton, R., Garrett, K.A., Percy, K., Hopkins, A., Davis, C., Karnosky, D.F., 2008. Impacts of global change on diseases of agricultural crops and forest trees. CAB Rev.: Perspect. Agric., Vet. Sci., Nutr. Nat. Resour. 3 (054), 1–15.

Congalton, R.G., 1991. A review of assessing the accuracy of classifcations of remotely sensed data. Remote Sens. Environ. 37 (1), 35–46.

Cuddihy, L.W., Stone, C.P., 1990. Alteration of Native Hawaiian Vegetation. Cooperative National Park Resources Studies Unit, Honolulu Hawaii, USA.

Dodd, R.S., Huberli, D., Douhovnikoff, V., Harnik, T.Y., Afzal-Rafi, Z., Garbelotto, M., 2005. Is variation in susceptibility to Phytophthora ramorum correlated with population genetic structure in coast live oak (Quercus agrifolia)? New Phytol. 165 (1), 203–214.

Frankel, S.J., Palmieri, K.M., 2014. Sudden Oak Death, Phytophthora ramorum: apersistent threat to oaks and other tree species. J. Int. Oak Soc. 25, 43–56.

Friday, J.B., Herbert, D., 2006. Metrosideros polymorpha (‘ohi‘a). In: Elevitch, C.R. (Ed.), Traditional Trees of Pacifc Islands. Permanent Agriculture Resources, Ho lualoa, HI, pp. 465–490 <http://www.traditionaltree.org>.

Giambelluca, T.W., Chen, Q., Frazier, A.G., Price, J.P., Chen, Y.L., Chu, P.S., Eischeid, J. K., Delparte, D., 2013. Online rainfall atlas of Hawai‘i. Bull. Am. Meteorol. Soc. 94 (3), 313–316 rainfall.geography.hawaii.edu (accessed March 2, 2016).

Gon, S.M., 2012. E K� ¯ hi‘a Lehua: Culture signifcance of a dominant Hawaiian u i Ka ‘Otree. Retrieved from: <https://www.scribd.com/doc/104237321/E-Ku-i-Ka-Ohia-Lehua-Culture-signifcance-of-a-dominant-Hawaiian-tree>, (accessed March 2, 2016).

Gon, S.M., Allison, A., Cannarella, R.J., Jacobi, J.D., Kaneshiro, K.Y., Kido, M.H., et al., 2006. A GAP Analysis of Hawaii: Final Report. US Department of the Interior. US Geological Survey, Washington, DC.

Hicke, J.A., Allen, C.D., Desai, A.R., Dietze, M.C., Hall, R.J., Kashian, D.M., Moore, D., Raffa, K.F., Sturrock, R.N., Vogelmann, J., 2012. Effects of biotic disturbances on forest carbon cycling in the United States and Canada. Glob. Change Biol. 18 (1), 7–34.

Hinds, T.E., 1972. Ceratocystis canker of Aspen. Phytopathology 62 (2), 213–220. Hosmer, R.S., 1911. The Part Played by the Forest in Conservation. The Hawaiian

Forester and Agriculturist 8:1, Honolulu, HI. Hodges, C.S., Adee, K.T., Stein, J.D., Wood, H.B., Doty, R.D., 1986. Decline of Ohia

(Metrosideros Polymorpha) in Hawaii: A Review. Gen. Tech. Rep. PSW-86. Pacifc Southwest Forest and Range Experiment Station, Forest Service, U.S. Department of Agriculture, Berkeley, CA.

Hughes, R.F., Denslow, J.S., 2005. Invasion by a N2-fxing tree alters function and structure in wet lowland forests of Hawaii. Ecol. Appl. 15 (5), 1615–1628.

Hughes, R.F., Asner, G.P., Mascaro, J., Uowolo, A., Baldwin, J., 2014. Carbon storage landscapes of lowland Hawaii: the role of native and invasive species through space and time. Ecol. Appl. 24 (4), 716–731.

Juzwik, J., Appel, D.N., MacDonald, W.L., Burks, S., 2011. Challenges and successes in managing oak wilt in the United States. Plant Dis. 95 (8), 888–900.

Keith, L.M., Hughes, R.F., Sugiyama, L.S., Heller, W.P., Bushe, B.C., Friday, J.B., 2015. First report of Ceratocystis wilt on ‘Ohi‘a. Plant Dis. 99 (9), 1276.

Landis, J.R., Koch, G.G., 1977. The measurement of observer agreement for categorical data. Biometrics, 159–174.

Little Jr., E.L., Skolmen, R.G., 1989. Common Forest trees of Hawaii (Native and Introduced). USDA Agriculture Handbook No. 679, (Washington) 679 DC..

92 L.A. Mortenson et al. / Forest Ecology and Management 377 (2016) 83–92

Lugo, A.E., Scatena, F.N., 1996. Background and catastrophic tree mortality in tropical moist, wet, and rain forests. Biotropica, 585–599.

Marin, M., Castro, B., Gaitan, A., Preisig, O., Wingfeld, B.D., Wingfeld, M.J., 2003. Relationships of Ceratocystis fimbriata isolates from Colombian coffee-growing regions based on molecular data and pathogenicity. J. Phytopathol. 151, 395–405.

Mascaro, J., Becklund, K.K., Hughes, R.F., Schnitzer, S.A., 2008. Limited native plant regeneration in novel, exotic-dominated forests on Hawai‘i. For. Ecol. Manage. 256 (4), 593–606.

McPherson, B.A., Mori, S.R., Wood, D.L., Kelly, M., Storer, A.J., Svihra, P., Standiford, R. B., 2010. Responses of oaks and tanoaks to the sudden oak death pathogen after 8 y of monitoring in two coastal California forests. For. Ecol. Manage. 259 (12), 2248–2255.

Miller, S.E., Eldredge, L.G., 1996. Numbers of Hawaiian species: supplement 1. Bishop Museum Occas. Pap. 45, 8–17, Honolulu, HI. Retrieved from: <http://hbs. bishopmuseum.org/pubs-online/pdf/op45p8-17.pdf> (accessed May 26, 2016).

Mitchell, C., Ogura, C., Meadows, D., Kane, A., Strommer, L., Fretz, S., Leonard, D., McClung, A., 2005. Hawaii’s Comprehensive Wildlife Conservation Strategy. Hawaii Department of Land and Natural Resources, Honolulu, HI, Retrieved from: <http://dlnr.hawaii.gov/wildlife/fles/2013/09/CWCS-Full-Document-2005Oct01.pdf> (accessed May 26, 2016).

Moore, R.B., Trusdell, F.A., 1991. Geologic Map of the Lower East Rift Zone of Kilauea Volcano, Hawaii. US Geological Survey Map I-2225.

Mueller-Dombois, D., 1987. Natural dieback in forests. Bioscience 37 (8), 575–583. Mueller-Dombois, D., Jacobi, J.D., Boehmer, H.J., Price, J., 2013. ‘O hi ‘a Lehua

Rainforest: Born Among Hawaiian Volcanoes, Evolved in Isolation. Friends of the Joseph Rock Herbarium, Honolulu, Hawaii, p. 269.

Mueller-Dombois, D., Fosberg, F.R., 1998. Vegetation of the Tropical Pacifc Islands. Springer, New York, p. 733.

Nelson, R., 1965. Record of Forest Plantings in Hawaii. US Forest Service, Pacifc Southwest Forest and Range Experiment Station, Forest Research Bulletin # 1.

Ocasio-Morales, R.G., Tsopelas, P., Harrington, T.C., 2007. Origin of Ceratocystis platani on native Platanus orientalis in Greece and its impact on natural forests. Plant Dis. 91 (7), 901–904.

Oblinger, B., Liu, Z., Bulaon, B., Fischer, L., 2012. Simple methods for mapping of tree mortality using readily available imagery: a case study from high meadows in the Lake Tahoe basin. Presented at 2012 Tahoe Science Conference. Retrieved from <http://tahoescience.org/events/conferences/conference-presentations/>.

Orwig, D.A., 2002. Ecosystem to regional impacts of introduced pests and pathogens: historical context, questions and issues. J. Biogeogr. 29 (10–11), 1471–1474.

Percy, D.M., Garver, A.M., Wagner, W.L., James, H.F., Cunningham, C.W., Miller, S.E., et al., 2008. Progressive island colonization and ancient origin of Hawaiian Metrosideros (Myrtaceae). Proc. Roy. Soc. B 275, 1479–1490.

Pontius Jr., R.G., Millones, M., 2011. Death to Kappa: birth of quantity disagreement and allocation disagreement for accuracy assessment. Int. J. Remote Sens. 32 (15), 4407–4429.

Pontius Jr., R.G., Santacruz, A., 2014. Quantity, exchange, and shift components of difference in a square contingency table. Int. J. Remote Sens. 35 (21), 7543–7554.

Poussio, G.B., Kazmi, M.R., Akem, C., Fateh, F.S., 2010. First record of Ceratocystis fmbriata associated with shisham (Dalbergia sissoo) decline in Pakistan. Australas. Plant Disease Notes 5 (1), 63–65.

Raupp, M.J., Cumming, A.B., Raupp, E.C., 2006. Street tree diversity in eastern North America and its potential for tree loss to exotic borers. Arboriculture Urban Forestry 32 (6), 297.

Rizzo, D.M., Garbelotto, M., 2003. Sudden oak death: endangering California and Oregon forest ecosystems. Front. Ecol. Environ. 1 (4), 197–204.

Roux, J., Wingfeld, M.J., Bouillet, J.-P., Wingfeld, B.D., Alfenas, A.C., 2000. A serious new wilt disease of Eucalyptus caused by Ceratocystis fimbriata in Central Africa. Forest Pathol. 30, 175–184.

Roux, J., Wingfeld, M.J., 2009. Ceratocystis species: emerging pathogens of non-native plantation Eucalyptus and Acacia species. Southern For.: J. For. Sci. 71 (2), 115–120.

Schulten, J.R., Cole, T.C., Cordell, S., Publico, K., Ostertag, R., Enoka, J.E., Michaud, J.D., 2014. Persistence of native trees in an invaded Hawaiian lowland wet forest: experimental evaluation of light and water constraints. Pac. Sci. 68 (2), 267– 285.

Shearer, B.L., Crane, C.E., Fairman, R.G., Dunne, C.P., 2009. Ecosystem dynamics altered by pathogen-mediated changes following invasion of Banksia woodland and Eucalyptus marginata forest biomes of south-western Australia by Phytophthora cinnamomi. Australas. Plant Pathol. 38 (4), 417–436.

Silva, A.C., Andrade, P.M., Alfenas, A.C., Graça, R.N., Cannon, P., Hauff, R., Ferreira, D. C., Mori, S., 2014. Virulence and Impact of Brazilian Strains of Puccinia psidii on Hawaiian‘ohi‘a (Metrosideros polymorpha). Pacifc Sci. 68 (1), 47–56.

Stack, R.W., McBride, D.K., Lamey, H.A., 2011. Dutch Elm Disease. North Dakota State Agriculture and University Extension <https://ndsu.edu/repository/ bitstream/handle/10365/9175/PP_324_1996.pdf?sequence=1>, PP 324 (accessed October 27, 2015).

Stacy, E.A., Johansen, J.B., Sakishima, T., Price, D.K., Pillon, Y., 2014. Incipient radiation within the dominant Hawaiian tree Meterosideros polymorpha. Heredity 113, 334–342.

Tarigan, M., Roux, J., Van, Wyk M., Tjahjono, B., Wingfeld, M.J., 2011. A new wilt and die-back disease of Acacia mangium associated with Ceratocystis manginecans and C. acaciivora sp. nov. in Indonesia. South Afr. J. Bot. 77 (2), 292–304.

Thu, P.Q., Qynh, D.N., Dell, B., 2012. Ceratocystis sp. causes crown wilt of Acacia spp. planted in some ecological zones of Vietnam. In: Proceeding of International Conference on The Impacts of Climate Change to Forest Pests and Diseases in The Tropics, 8–10 October, Yogyakarta, Indonesia, pp. 38–44.

Uchida, J., Zhong, S., Killgore, E., 2006. First report of a rust disease on Ohia caused by Puccinia psidii in Hawaii. Plant Dis. 90 (4), 524–524.

USDA Forest Service, 2013. Field Instructions for the Annual Inventory of California, Oregon and Washington. Pacifc Northwest Research Station, Portland, OR.

Wagner, W.L., Herbst, D.R., Sohmer, S.H., 1999. Manual of the Flowering Plants of Hawai‘i, Revised ed. University of Hawaii Press, Bishop Museum Press, Honolulu, Hawaii.

Walters, J.W., 1992. Dutch elm disease. In: Hutchinson, J.G. (Ed.), Northern Hardwood Notes. U.S. Dept. of Agriculture, Forest Service, North Central Forest Experiment Station. 7.03, St. Paul, MN.

Zimmerman, N., Hughes, R.F., Cordell, S., Hart, P., Chang, H.K., Perez, D., Like, R. K., Ostertag, R., 2008. Patterns of primary succession of native and introduced plants in lowland wet forests in eastern Hawai‘i. Biotropica 40 (3), 277–284.